Skip to content

Latest commit

 

History

History
7 lines (6 loc) · 892 Bytes

File metadata and controls

7 lines (6 loc) · 892 Bytes

Transformer models categorized by architecture type

Category Description Famous Open-Source Models
Encoder-Only Understanding-focused models for tasks like classification, QA (Extractive), embeddings BERT, RoBERTa, ALBERT, DistilBERT, ELECTRA, DeBERTa, ERNIE
Decoder-Only Generation-focused models for text completion and open-ended output GPT (1-4), GPT-Neo, GPT-J, GPT-NeoX, LLaMA, Mistral, Phi, Gemma
Encoder-Decoder (Seq2Seq) Input-to-output models for translation, summarization, etc. T5, mT5, FLAN-T5, BART, mBART, UL2, MarianMT, ByT5